Tomlin (1986) buruzki Perkin (1980): "... while Perkin's sample, like Greenberg's, is useful for exploring correlations among variables of interest, it is not useful for estimating the frequency of basic constituent orders..."

Atzokoan amaitzen genuen azpimarrátuz ondoko reflexioa ganik Tomlin (1986):

... that size alone cannot guarantee sample representativeness . [Tomlin, 1986:19-20]

Gaurkoan, ordea, itzuliko gara ki beste lagin txiki bat (50 hizkuntza) zein,  Tomlin-ek dioenez, izan litekén representatiboa on diversitate linguistikoa (hartzen baititú hizkuntza diferenteak tikan munduko lokalizazio diferenteak), baina ez da representagarria respektu proportzioak on hizkuntzak an lokalizazio diferenteak. Horrela, hartzen ditú proportzionalki oso hizkuntza gutxi artén Niger-Congo hizkuntzak, bitárten hartzen dituén erlatiboki gehiegi an Iparramerika. Ikus an hitzak ganik Tomlin (1986):

Although Perkin's sample may be representative of the diversity of languages ​​(and cultures), it is not representative of the universe of natural languages . For instance, Perkin's sample includes a single language for all of Niger-Congo , 2% of his final sample , even though there are more than 1000 Niger-Congo languages , or approximately 20% of the Voegelin and Voegelin total. Similarly, the sample contains 17 languages ​​from the North American continent (including Central America), 34% of the final sample , yet languages ​​from that area represent at most some 8% of the estimated world total . Thus, while Perkin's sample, like Greenberg's , is useful for exploring correlations among variables of interest , it is not useful for estimating the frequency of basic constituent orders in the universe of natural languages. [Tomlin, 1986:20]

Hortaz, datuak izan ahal dira produktiboak noiz egin zenbait analisi, baina ez noiz helburua dén lortzea irudi razonable bat on munduko hitz-ordenen banaketa reala. [2754]